In this paper a fast and innovative three-dimensional vision system, having high resolution in the surface reconstruction, is discussed. It is based on a triangulation 3D laser scanner with a linear beam shape. The high precision (few microns) is guaranteed by very small laser line width, small camera pixel-size and proper optical properties of the Telecentric Lens. The entire system has been tested on two kinds of sample objects such as a 20 cent coin and a set of precision drilling tools. The main purpose of this work is the detection and reconstruction of the 3D surface of tiny objects and the measurement of their surface defects with high accuracy. Furthermore the occlusion problem is faced and solved by properly handling the camera-laser setup. Experimental tests prove the high precision of the system that can reach a resolution of 15 ?m. © 2013 IEEE.

In this paper, an accurate range sensor for the three-dimensional reconstruction of environments is designed and developed. Following the principles of laser profilometry, the device exploits a set of optical transmitters able to project a laser line on the environment. A high-resolution and high-frame-rate camera assisted by a telecentric lens collects the laser light reflected by a parabolic mirror, whose shape is designed ad hoc to achieve a maximum measurement error of 10 mm when the target is placed 3 m away from the laser source. Measurements are derived by means of an analytical model, whose parameters are estimated during a preliminary calibration phase. Geometrical parameters, analytical modeling and image processing steps are validated through several experiments, which indicate the capability of the proposed device to recover the shape of a target with high accuracy. Experimental measurements show Gaussian statistics, having standard deviation of 1.74 mm within the measurable range. Results prove that the presented range sensor is a good candidate for environmental inspections and measurements.

One of the first tasks executed by a vision system made of fixed cameras is the background (BG) subtraction and a particularly challenging context for real time applications is the athletic one because of illumination changes, moving objects and cluttered scenes. The aim of this work is to extract a BG model based on statistical likelihood able to process color filter array (CFA) images taking into account the intrinsic variance of each gray level of the sensor, named Likelihood Bayer Background (LBB). The BG model should be not so computationally complex while highly responsive to extract a robust foreground. Moreover, the mathematical operations used in the formulation should be parallelizable, working on image patches, and computationally efficient, exploiting the dynamics of a pixel within its integer range. Both simulations and experiments on real video sequences demonstrate that this BG model approach shows great performances and robustness during the real time processing of scenes extracted from a soccer match.

In this article, an accurate method for the registration of point clouds returned by a 3D rangefinder is presented. The method modifies the well-known iterative closest point (ICP) algorithm by introducing the concept of deletion mask. This term is defined starting from virtual scans of the reconstructed surfaces and using inconsistencies between measurements. In this way, spatial regions of implicit ambiguities, due to edge effects or systematical errors of the rangefinder, are automatically found. Several experiments are performed to compare the proposed method with three ICP variants. Results prove the capability of deletion masks to aid the point cloud registration, lowering the errors of the other ICP variants, regardless the presence of artifacts caused by small changes of the sensor view-point and changes of the environment.

Camera triggering represents an essential step for synchronizing artificial vision systems (AVSs) and can affect the quality of acquired images. In fact, a proper trigger signal is mandatory to synchronize in time both stand alone or multiple cameras covering large environments. In addition, indoor environments with artificial light sources can induce flickering noise in captured frames, affecting the performance of the algorithms that are usually executed to interpret a scene or perform various tasks. In this letter, we describe the design of an embedded system for camera triggering that can be employed to deal with such issues and remove flickering noise while capturing an image with the highest possible dynamic range. Experiments on real data show how the proposed trigger can be effectively employed in AVSs.

Background (BG) modelling is a key task in every computer vision system (CVS) independently of the final purpose for which it is designed. Even if many BG approaches exist (for example Mixture of Gaussians or Eigenbackground), they can not efficiently process real time videos due to the model complexity and to the high throughput of the video flux. One of the most challenging real time applications is the athletic scene processing, because, in this context, there are many critical aspects for defining a BG model: no a-priori knowledge of the static scene, sudden illumination changes and many moving objects that slow down the upgrade phase. The aim of this work is to provide an adaptive BG model able to deal with high frame rate videos (>= 100 fps) in real time processing, and suitable for smart cameras embedding, finding a good compromise between the model complexity and its responsiveness. Real experiments demonstrate that this BG model approach shows great performances and robustness during the real time processing of athletic video frames, up to 100 fps. Copyright 2014 ACM.

In this paper, we propose an embedded vision system based on laser profilometry able to get the pose of a vehicle and its relative displacements with reference to the constitutive media of a structured environment. Fundamental equations for laser triangulation are developed and encoded for their actual implementation on an embedded system. It is made of a laser source that projects a line-shaped beam onto the environment and an on-chip camera able to frame the laser light. Images are then sent to the inexpensive Raspberry Pi onboard computer, which is responsible for processing tasks. For the first time, laser profilometry is coupled with the correlation of laser signatures on a low-cost and low-resource processing board for vehicle localization purposes. Several validation tests of the proposed sensor have proven the effectiveness of the system with respect to commercially available sensors such as inductive sensors and standard odometers, which fail when the vehicle crosses path interceptions or its wheels undergo unavoidable slippages. Moreover, further comparisons with other vision-based techniques have also proven the good performances of this embedded system for real-time localization of vehicles.

In this paper, an approach based on the analysis of variance (ANOVA) for the extraction of crop marks from aerial images is improved by means of preliminary analyses and semantic processing of the extracted objects. The paper falls in the field of digitalization of images for archaeology, assisting expert users in the detection of un-excavated sites. The methodology is improved by a preliminary analysis of local curvatures, able to determine the most suitable direction for the ANOVA formulation. Then, a semantic processing, based on the knowledge of the shape of the target wide line, is performed to delete false positive detections. Sample analyses are always performed on actual images and prove the capability of the method to discriminate the most significant marks, aiding archaeologists in the analysis of huge amount of data.

In this paper we present a reliable method to derive the differences between indoor environments using the comparison of high-resolution range images. Samples belonging to different acquisitions are firstly reduced preserving the topology of the scenes and then registered in the same system of reference through an iterative least-squares algorithm, aided by a deletion mask, whose assignment is the removal of implicit errors due to the different points of view of each orthographic acquisition. Finally the analysis of the exact range measures returns an intuitive difference map that allows the fast detection of the positions of the altered regions within the scenes. Numerical experiments are presented to prove the capability of the method for the comparison of scenes regardless the resolution of the sensor and the input noise level of such measurements. © 2013 IEEE.

This paper considers the problem of detecting archaeological traces in digital aerial images by analyzing the pixel variance over regions around selected points. In order to decide if a point belongs to an archaeological trace or not, its surrounding regions are considered. The one-way ANalysis Of VAriance (ANOVA) is applied several times to detect the differences among these regions; in particular the expected shape of the mark to be detected is used in each region. Furthermore, an effect size parameter is defined by comparing the statistics of these regions with the statistics of the entire population in order to measure how strongly the trace is appreciable. Experiments on synthetic and real images demonstrate the effectiveness of the proposed approach with respect to some state-of-the-art methodologies.

This paper presents a complete framework aimed to nondestructive inspection of composite materials. Starting from the acquisition, performed with lock-in thermography, the method flows through a set of consecutive blocks of data processing: input enhancement, feature extraction, classification and defect detection. Experimental results prove the capability of the presented methodology to detect the presence of defects underneath the surface of a calibrated specimen made of Glass Fiber Reinforced Polymer (GFRP). Results are also compared with those obtained by other techniques, based on different features and unsupervised learning methods. The comparison further proves that the proposed methodology is able to reduce the number of false positives, while ensuring the exact detection of subsurface defects.

In recent years sport video research has gained a steady interest among the scientific community. The large amount of video data available from broadcast transmissions and from dedicated camera setups, and the need of extracting meaningful information from data, pose significant research challenges. Hence, computer vision and machine learning are essential for enabling automated or semi-automated processing of big data in sports. Although sports are diverse enough to present unique challenges on their own, most of them share the need to identify active entities such as ball or players. In this paper, an innovative deep learning approach to the identification of the ball in tennis context is presented. The work exploits the potential of a convolutional neural network classifier to decide whether a ball is being observed in a single frame, overcoming the typical issues that can occur dealing with classical approaches on long video sequences (e.g. illumination changes and flickering issues). Experiments on real data confirm the validity of the proposed approach that achieves 98.77% accuracy and suggest its implementation and integration at a larger scale in more complex vision systems.

In the field of NDT techniques for aeronautic components of composite materials, the development of automatic and robust approaches for defect detection is largely desirable for both safety and economic reasons. This paper introduces a novel methodology for the automatic analysis of thermal signals resulting from the application of pulsed thermography. Input thermal decays are processed by a proper FIR filter designed to reduce the measurement noise, and then modeled to represent both sound regions and defective ones. Output signals are thus fitted on an exponential model, which approximates thermal contrasts with three robust parameters. These features feed a decision forest, trained to detect discontinuities and characterize their depths. Several experiments on actual sample laminates have proven the increase of the classification performance of the proposed approach with respect to related ones in terms of the reduction of missing predictions of defective classes.

A high-resolution vision system for the inspection of drilling tools is presented. A triangulation-based laser scanner is used to extract a three-dimensional model of the target aimed to the fast detection and characterization of surface defects. The use of two orthogonal calibrated handlings allows the achievement of precisions of the order of few microns in the whole testing volume and the prevention of self-occlusions induced on the undercut surfaces of the tool. Point cloud registration is also derived analytically to increase to strength of the measurement scheme, whereas proper filters are used to delete samples whose quality is below a reference threshold. Experimental tests are performed on calibrated spheres and different-sized tools, proving the capability of the presented setup to entirely reconstruct complex targets with maximum absolute errors between the estimated distances and the corresponding nominal values below 12 mu m.

High resolution in distance (range) measurements can be achieved by means of accurate instrumentations and precise analytical models. This paper reports an improvement in the estimation of distance measurements performed by an omnidirectional range sensor already presented in literature. This sensor exploits the principle of laser triangulation, together with the advantages brought by catadioptric systems, which allow the reduction of the sensor size without decreasing the resolution. Starting from a known analytical model in two dimensions (2D), the paper shows the development of a fully 3D formulation where all initial constrains are removed to gain in measurement accuracy. Specifically, the ray projection problem is solved by considering that both the emitter and the receiver have general poses in a global system of coordinates. Calibration is thus made to estimate their poses and compensate for any misalignment with respect to the 2D approximation. Results prove an increase in the measurement accuracy due to the more general formulation of the problem, with a remarkable decrease of the uncertainty.

Plant phenotyping, that is, the quantitative assessment of plant traits including growth, morphology, physiology, and yield, is a critical aspect towards efficient and effective crop management. Currently, plant phenotyping is a manually intensive and time consuming process, which involves human operators making measurements in the field, based on visual estimates or using hand-held devices. In this work, methods for automated grapevine phenotyping are developed, aiming to canopy volume estimation and bunch detection and counting. It is demonstrated that both measurements can be effectively performed in the field using a consumer-grade depth camera mounted on-board an agricultural vehicle. First, a dense 3D map of the grapevine row, augmented with its color appearance, is generated, based on infrared stereo reconstruction. Then, different computational geometry methods are applied and evaluated for plant per plant volume estimation. The proposed methods are validated through field tests performed in a commercial vineyard in Switzerland. It is shown that different automatic methods lead to different canopy volume estimates meaning that new standard methods and procedures need to be defined and established. Four deep learning frameworks, namely the AlexNet, the VGG16, the VGG19 and the GoogLeNet, are also implemented and compared to segment visual images acquired by the RGB-D sensor into multiple classes and recognize grape bunches. Field tests are presented showing that, despite the poor quality of the input images, the proposed methods are able to correctly detect fruits, with a maximum accuracy of 91.52%, obtained by the VGG19 deep neural network.

Pulsed thermography has been used for many years to investigate the presence of subsurface defects in composite materials for aeronautics. Several methods have been proposed but only few of them include a complete automated approach for the effective defect characterization. This paper presents a novel method which approximates the thermal decays on the laminate surface, induced by a short heat pulse, by means of an exponential model in three unknowns (model parameters), estimated in the least squares sense. These parameters are discriminant and noise-insensitive features used to feed several classifiers, which are trained to label possible defects according to their depths. Experimental tests have been performed on a carbon-fiber reinforced polymer (CFRP) laminate having four inclusions of known properties. The comparative analysis of the proposed classifiers has demonstrated that the best results are achieved by a decision forest made of 30 trees. In this case the mean values of standard and balanced accuracies reach 99.47% and 86.9%, whereas precision and recall are 89.87% and 73.67%, respectively.

This paper describes a complete method for monitoring indoor environments. Three-dimensional (3D) point clouds are first acquired from the environment under investigation by means of a laser range scanner in order to obtain several 3D models to be compared. Input datasets are thus registered each other exploiting a reliable variant of the iterative closest point algorithm (ICP) assisted by the use of deletion masks. These terms work in cooperation with the resampling of the model surfaces to reduce significantly the errors in the estimation of the registration parameters. Once datasets are registered, deformation maps are displayed to help the user to detect changes within the environment. Deletion masks are again used to filter measurement artifacts from the comparison, thus highlighting only the actual alterations of the environment. Several experiments are performed for the analysis of an indoor environment, proving the capability of the proposed method to reliably estimate the presence of alterations.

This paper tackles the problem of people re-identification by using soft biometrics features. The method works on RGB-D data (color point clouds) to determine the best matching among a database of possible users. For each subject under testing, skeletal information in three-dimensions is used to regularize the pose and to create a skeleton standard posture (SSP). A partition grid, whose sizes depend on the SSP, groups the samples of the point cloud accordingly to their position. Every group is then studied to build the person signature. The same grid is then used for the other subjects of the database to preserve information about possible shape differences among users. The effectiveness of this novel method has been tested on three public datasets. Numerical experiments demonstrate an improvement of results with reference to the current state-of-the-art, with recognition rates of 97.84% (on a partition of BIWI RGBD-ID), 61.97% (KinectREID) and 89.71% (RGBD-ID), respectively.

In this paper we present a natural humancomputer interface based on gesture recognition. The principal aimis to study how different personalized gestures, defined by users,can be represented in terms of features and can be modelled byclassification approaches in order to obtain the best performancesin gesture recognition. Ten different gestures involving themovement of the left arm are performed by different users.Different classification methodologies (SVM, HMM, NN, and DTW) arecompared and their performances and limitations are discussed. Anensemble of classifiers is proposed to produce more favorableresults compared to those of a single classifier system. Theproblems concerning different lengths of gesture executions,variability in their representations, generalization ability ofthe classifiers have been analyzed and a valuable insight inpossible recommendation is provided.

This paper analyzes with a new perspective the recent state of-the-art on gesture recognition approaches that exploit both RGB and depth data (RGB-D images). The most relevant papers have been analyzed to point out which features and classifiers best work with depth data, if these fundamentals are specifically designed to process RGB-D images and, above all, how depth information can improve gesture recognition beyond the limit of standard approaches based on solely color images. Papers have been deeply reviewed finding the relation between gesture complexity and features/methodologies suitability. Different types of gestures are discussed, focusing attention on the kind of datasets (public or private) used to compare results, in order to understand weather they provide a good representation of actual challenging problems, such as: gesture segmentation, idle gesture recognition, and length gesture invariance. Finally the paper discusses on the current open problems and highlights the future directions of research in the field of processing of RGB-D data for gesture recognition.

Non-destructive testing is essential for the thorough assessment of production processes of complex materials, such as composites. This paper presents a complete algorithm to detect subsurface defects, e.g. extended delaminations or local resin pockets, by comparing the outputs produced by lock-in thermography for the inspection of master pristine samples and the current ones under testing. The use of lock-in thermography produces amplitude and phase maps. Focusing on amplitudes, dataset are first made comparable in both magnitude spans and spatial positions exploiting image normalization and alignment. Then local patches in actual correspondence are cross-correlated to further improve their alignment and estimate a similarity measurement. Differences in thermal behaviors detected by the proposed processing underlie subsurface defects. These outcomes have been also proven by experimental investigations performed on a carbon fiber reinforced polymer (CFRP) T-joint.